Search Results (63)

Falls remain a critical safety concern in surveillance settings, yet monocular RGB methods often degrade in multi-person scenes with occlusion and loss of three-dimensional cues. This study proposes a neuro-symbolic framework that restores physically interpretable depth proxies from monocular video and fuses them with skeleton-based spatio-temporal inference for robust fall detection. The pipeline estimates per-frame depth and 2D skeletons, recovers world coordinates for key joints, and derives absolute neck height and vertical descent rate for rule-based adjudication, while a neural method operates on joint trajectories; final decisions combine both streams with a logical policy and short-horizon temporal consistency. Experiments in a realistic indoor testbed with multi-person activity compare three configurations—neural, symbolic, and fused. The fused neuro-symbolic method achieved an accuracy of 0.88 and an F1 score of 0.76 on the real surveillance test set, outperforming the neural method alone (accuracy 0.81, F1 0.64) and the symbolic method alone (accuracy 0.77, F1 0.35). Gains arise from complementary error profiles: depth-derived, rule-based cues suppress spurious positives on non-fall frames, while the neural stream recovers true falls near rule boundaries. These findings indicate that integrating monocular depth proxies with interpretable rules improves reliability without additional sensors, supporting deployment in complex, multi-person surveillance environments. Full article

Accurate and timely diagnosis of rice leaf diseases is crucial for smart agriculture leveraging vision sensors. However, existing lightweight convolutional neural networks (CNNs) often struggle in complex field environments, where small lesions, cluttered backgrounds, and varying illumination complicate recognition. This paper presents I-GhostNetV3, an incrementally improved GhostNetV3-based network for RGB rice leaf disease recognition. I-GhostNetV3 introduces two modular enhancements with controlled overhead: (1) Adaptive Parallel Attention (APA), which integrates edge-guided spatial and channel cues and is selectively inserted to enhance lesion-related representations (at the cost of additional computation), and (2) Fusion Coordinate-Channel Attention (FCCA), a near-neutral SE replacement that enables efficient spatial–channel feature fusion to suppress background interference. Experiments on the Rice Leaf Bacterial and Fungal Disease (RLBF) dataset show that I-GhostNetV3 achieves 90.02% Top-1 accuracy with 1.831 million parameters and 248.694 million FLOPs, outperforming MobileNetV2 and EfficientNet-B0 under our experimental setup while remaining compact relative to the original GhostNetV3. In addition, evaluation on PlantVillage-Corn serves as a supplementary transfer sanity check; further validation on independent real-field target domains and on-device profiling will be explored in future work. These results indicate that I-GhostNetV3 is a promising efficient backbone for future edge deployment in precision agriculture. Full article

The fermentation process of oolong tea is a critical step in shaping its quality and flavor profile. In this study, the fermentation degree of Anxi Tieguanyin oolong tea was assessed using image and hyperspectral features. Machine learning algorithms, including Support Vector Machine (SVM), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU), were employed to develop models based on both single-source features and multi-source fused features. First, color and texture features were extracted from RGB images and then processed through Pearson correlation-based feature selection and Principal Component Analysis (PCA) for dimensionality reduction. For the hyperspectral data, preprocessing was conducted using Normalization (Nor) and Standard Normal Variate (SNV), followed by feature selection and dimensionality reduction with Competitive Adaptive Reweighted Sampling (CARS), Successive Projections Algorithm (SPA), and PCA. We then performed mid-level fusion on the two feature sets and selected the most relevant features using L1 regularization for the final modeling stage. Finally, SHapley Additive exPlanations (SHAP) analysis was conducted on the optimal models to reveal key features from both hyperspectral bands and image data. The results indicated that models based on single features achieved test set accuracies of 68.06% to 87.50%, while models based on data fusion achieved 77.78% to 94.44%. Specifically, the Pearson+Nor-SPA+L1+SVM fusion model achieved the highest accuracy of 94.44%. This demonstrates that data feature fusion enables a more comprehensive characterization of the fermentation process, significantly improving model accuracy. SHAP analysis revealed that the hyperspectral bands at 967, 942, 814, 784, 781, 503, 413, and 416 nm, along with the image features H_σ and H, played the most crucial roles in distinguishing tea fermentation stages. These findings provide a scientific basis for assessing the fermentation degree of Tieguanyin oolong tea and support the development of intelligent detection systems. Full article

Accurate evaluation of demand response (DR) potential at the individual user level is critical for the effective implementation and optimization of demand response programs. However, existing data-driven methods often suffer from insufficient feature representation, limited characterization of load profile dynamics, and ineffective fusion of heterogeneous features, leading to suboptimal evaluation performance. To address these challenges, this paper proposes a novel demand response potential evaluation method based on multivariate heterogeneous features and a Stacking-based ensemble mechanism. First, multidimensional indicator features are extracted from historical electricity consumption data and external factors (e.g., weather, time-of-use pricing), capturing load shape, variability, and correlation characteristics. Second, to enrich the information space and preserve temporal dynamics, typical daily load profiles are transformed into two-dimensional image features using the Gramian Angular Difference Field (GADF), the Markov Transition Field (MTF), and an Improved Recurrence Plot (IRP), which are then fused into a single RGB image. Third, a differentiated modeling strategy is adopted: scalar indicator features are processed by classical machine learning models (Support Vector Machine, Random Forest, XGBoost), while image features are fed into a deep convolutional neural network (SE-ResNet-20). Finally, a Stacking ensemble learning framework is employed to intelligently integrate the outputs of base learners, with a Decision Tree as the meta-learner, thereby enhancing overall evaluation accuracy and robustness. Experimental results on a real-world dataset demonstrate that the proposed method achieves superior performance compared to individual models and conventional fusion approaches, effectively leveraging both structured indicators and unstructured image representations for high-precision demand response potential evaluation. Full article

Fomes fomentarius (L.) Fr. sensu lato is a common, widespread polypore and a pathological decayer in many hosts such as poplar, beech, and birch. It is either regarded as a single species, a species complex, or displaying a significant intraspecific variability. Limits between populations are fuzzy, and local differences have been mainly related to the current distribution of preferred hosts. The aim of the work was to test an imaging technique (RGB profiling) of cultures’ macromorphology on Petri plates to implement the traditional growth profiles of pure cultures in order to point out differences between strains from different European regions, hosts, and climates. Growth rates at 24 °C and 30 °C poorly segregated strains based on the origin, whereas there is a marked difference at 15 °C between strains from oceanic climates and continental climates. K-means clustering of RGB profiles also marked a difference at 15 °C between Central/North European strains and the Italian strains, although this variability gradually attenuates by increasing temperature. The combined approach, including a radial growth measuring and RGB profiling, successfully pointed out the intraspecific diversity in F. fomentarius, suggesting local adaptations. This study contributes to establishing a methodology to investigate the ecotype concept in polypores. Full article

Due to the shortage of rural labor and an increasingly aging population, promoting the mechanized plucking of bulk tea and improving plucking efficiency have become urgent problems for tea plantations. Previous bulk tea plucking machines have not fully adapted to tea plantations in hilly areas, necessitating enhancements in the performance of cutter profiling. In this paper, we present an automatic cutter profiling method based on an RGB-D camera, which utilizes the depth information of bulk tea shoots to tackle the issues mentioned above. Specifically, we use improved super-green features and the Otsu method to detect and segment the shoots from the RGB images of the tea canopy taken from different lighting conditions. Furthermore, the cutting pose based on the depth value of the tea shoots can be generated as a basis for cutter profiling. Lastly, the profiling task is completed by the upper computer controlling motors to adjust the cutter pose. Field tests were conducted in the tea plantation to verify the proposed profiling method’s effectiveness. The average bud and leaf integrity rate, leakage rate, loss rate, tea making rate, and qualified rate were 81.2%, 0.91%, 0.66%, and 90.4%, respectively. The results show that the developed algorithm can improve cutting pose calculation accuracy and that the harvested bulk tea shoots meet the requirements of machine plucking quality standards and the subsequent processing process. Full article

Driver monitoring systems benefit from fixation-related eye-tracking features, yet dedicated eye-tracking hardware is costly and difficult to integrate at scale. This study presents a practical software pipeline that extracts fixation-related features from conventional RGB video. Facial and pupil landmarks obtained with MediaPipe are denoised using a Kalman filter, fixation centers are identified with the OPTICS algorithm within a sliding window, and an affine normalization compensates for head motion and camera geometry. Fixation segments are derived from smoothed velocity profiles based on a moving average. Experiments with laptop camera recordings show that the combined Kalman and OPTICS pipeline reduces landmark jitter and yields more stable fixation centroids, while the affine normalization further improves referential pupil stability. The pipeline operates with minimal computational overhead and can be implemented as a software update in existing driver monitoring or advanced driver assistance systems. This work is a proof of concept that demonstrates feasibility in a low-cost RGB setting with a limited evaluation scope. Remaining challenges include sensitivity to lighting conditions and head motion that future work may address through near-infrared sensing, adaptive calibration, and broader validation across subjects, environments, and cameras. The extracted features are relevant for future studies on cognitive load and attention, although cognitive state inference is not validated here. Full article

Classifying brain tumour transcriptomic data is crucial for precision medicine but remains challenging due to high dimensionality and limited interpretability of conventional models. This study benchmarks three image-based deep learning approaches, DeepInsight, Fotomics, and a novel saliency-guided convolutional neural network (CNN), for transcriptomic classification. DeepInsight utilises dimensionality reduction to spatially arrange gene features, while Fotomics applies Fourier transforms to encode expression patterns into structured images. The proposed method transforms each single-cell gene expression profile into an RGB image using PCA, UMAP, or t-SNE, enabling CNNs such as ResNet to learn spatially organised molecular features. Gradient-based saliency maps are employed to highlight gene regions most influential in model predictions. Evaluation is conducted on two biologically and technologically different datasets: single-cell RNA-seq from glioblastoma GSM3828672 and bulk microarray data from medulloblastoma GSE85217. Outcomes demonstrate that image-based deep learning methods, particularly those incorporating saliency guidance, provide a robust and interpretable framework for uncovering biologically meaningful patterns in complex high-dimensional omics data. For instance, ResNet-18 achieved the highest accuracy of 97.25% on the GSE85217 dataset and 91.02% on GSM3828672, respectively, outperforming other baseline models across multiple metrics. Full article

Food quality and safety are paramount in preserving the culinary authenticity and cultural integrity of Chinese cuisine, characterized by intricate ingredient combinations, diverse cooking techniques (e.g., stir-frying, steaming, and braising), and region-specific flavor profiles. Traditional non-destructive detection methods often struggle with the unique challenges posed by Chinese dishes, including complex textural variations in staple foods (e.g., noodles, dumplings), layered seasoning compositions (e.g., soy sauce, Sichuan peppercorns), and oil-rich cooking media. This study pioneers a hyperspectral imaging framework enhanced with domain-specific deep learning algorithms (spatial–spectral convolutional networks with attention mechanisms) to address these challenges. Our approach effectively deciphers the subtle spectral fingerprints of Chinese-specific ingredients (e.g., fermented black beans, lotus root) and quantifies critical quality indicators, achieving an average classification accuracy of 97.8% across 15 major Chinese dish categories. Specifically, the model demonstrates high precision in quantifying chili oil content in Mapo Tofu with a Mean Absolute Error (MAE) of 0.43% w/w and assessing freshness gradients in Cantonese dim sum (Shrimp Har Gow) with a classification accuracy of 95.2% for three distinct freshness levels. This approach leverages the detailed spectral information provided by hyperspectral imaging to automate the classification and detection of Chinese dishes, significantly improving both the accuracy of image-based food classification by >15 percentage points compared to traditional RGB methods and enhancing food quality safety assessment. Full article

This study presents an approach to macroalgae monitoring using a cost-effective hyperspectral imaging (HSI) system and artificial intelligence (AI). Kelp beds are vital habitats and support nutrient cycling, making ongoing monitoring crucial amid environmental changes. HSI emerges as a powerful tool in this context, due to its ability to detect pigment-characteristic fingerprints that are often missed altogether by standard RGB cameras. Still, the high costs of these systems are a barrier to large-scale deployment for in situ monitoring. Here, we showcase the development of a cost-effective HSI setup that combines a GoPro camera with a continuous linear variable spectral bandpass filter. We empirically validate the operational capabilities through the analysis of two brown macroalgae, Fucus serratus and Fucus versiculosus, and two red macroalgae, Ceramium sp. and Vertebrata byssoides, in a controlled aquatic environment. Our HSI system successfully captured spectral information from the target species, which exhibit considerable similarity in morphology and spectral profile, making them difficult to differentiate using traditional RGB imaging. Using a one-dimensional convolutional neural network, we reached a high average classification precision, recall, and F1-score of 99.9%, 89.5%, and 94.4%, respectively, demonstrating the effectiveness of our custom low-cost HSI setup. This work paves the way to achieving large-scale and automated ecological monitoring. Full article

Remotely Piloted Aircraft (RPA) as sensor-carrying airborne platforms for indirect measurement of plant physical parameters has been discussed in the scientific community. The utilization of RGB sensors with photogrammetric data processing based on Structure-from-Motion (SfM) and Light Detection and Ranging (LiDAR) sensors for point cloud construction are applicable in this context and can yield high-quality results. In this sense, this study aimed to compare coffee plant height data obtained from RGB/SfM and LiDAR point clouds and to estimate soil compaction through penetration resistance in a coffee plantation located in Minas Gerais, Brazil. A Matrice 300 RTK RPA equipped with a Zenmuse L1 sensor was used, with RGB data processed in PIX4D software (version 4.5.6) and LiDAR data in DJI Terra software (version V4.4.6). Canopy Height Model (CHM) analysis and cross-sectional profile, together with correlation and statistical difference studies between the height data from the two sensors, were conducted to evaluate the RGB sensor’s capability to estimate coffee plant height compared to LiDAR data considered as reference. Based on the height data obtained by the two sensors, soil compaction in the coffee plantation was estimated through soil penetration resistance. The results demonstrated that both sensors provided dense point clouds from which plant height (R2 = 0.72, R = 0.85, and RMSE = 0.44) and soil penetration resistance (R2 = 0.87, R = 0.8346, and RMSE = 0.14 m) were accurately estimated, with no statistically significant differences determined between the analyzed sensor data. It is concluded, therefore, that the use of remote sensing technologies can be employed for accurate estimation of coffee plantation heights and soil compaction, emphasizing a potential pathway for reducing laborious manual field measurements. Full article

Spatio-temporal measurements of turbulent free surface flows remain challenging with in situ point methods. This study explores the application of an inexpensive depth-sensing RGB-D camera, the Intel^® RealSense™ D455, to capture detailed water surface measurements of a highly turbulent, self-aerated flow in the case of a stepped spillway. Ambient lighting conditions and various sensor settings, including configurations and parameters affecting data capture and quality, were assessed. A free surface profile was extracted from the 3D measurements and compared against phase detection conductivity probe (PDCP) and ultrasonic sensor (USS) measurements. Measurements in the non-aerated region were influenced by water transparency and a lack of detectable surface features, with flow depths consistently smaller than USS measurements (up to 32.5% less). Measurements in the clear water region also resulted in a “no data” region with holes in the depth map due to shiny reflections. In the aerated flow region, the camera effectively detected the dynamic water surface, with mean surface profiles close to characteristic depths measured with PDCP and within one standard deviation of the mean USS flow depths. The flow depths were within 10% of the USS depths and corresponded to depths with 80–90% air concentration levels obtained with the PDCP. Additionally, the depth camera successfully captured temporal fluctuations, allowing for the calculation of time-averaged entrapped air concentration profiles and dimensionless interface frequency distributions. This facilitated a direct comparison with PDCP and USS sensors, demonstrating that this camera sensor is a practical and cost-effective option for detecting free surfaces of high velocity, aerated, and dynamic flows in a stepped chute. Full article

To obtain fundamental information on the Tequila 100% agave Cristalino commercial samples were characterized in their different classes. For this purpose, 12 samples were chosen, defined as: G1 (aged; n = 3, or extra-aged; n = 3) and G2 (aged-Cristalino; n = 3 or extra-aged-Cristalino; n = 3). Analytical characterization was performed on these beverages, consisting of isotope ratio mass spectrometry, gas and liquid chromatography, UV-Vis spectroscopy, and color using digital image processing. The results corroborate that the chromatographic characterization (mg/100 mL A.A.)—higher alcohols (299.53 ± 46.56), methanol (212.02 ± 32.28), esters (26.02 ± 4.60), aldehydes (8.93 ± 4.61), and furfural (1.02 ± 0.56)—and isotopic characterization—δ¹³C_VPDB = −13.02 ± 0.35 ‰ and δ¹⁸O_VSMOW = 21.31 ± 1.33 ‰—do not present statistically significant differences (p > 0.05) between groups. From these techniques, it was possible to reinforce that isotopic ratios can provide information about that the ethanol of these alcoholic beverages come from Agave tequilana Weber blue variety and it is not affected in the filtration process. Based on the UV-Vis analysis, I₂₈₀ and I₃₆₅ were obtained, which were related to the presence of polyphenols and flavonoids—expressed as mg quercetin equivalents/L—only found in group 1. Due to the presence of flavonoids in aged beverages, the oxidation process results in the formation of an amber color, which can be measured by an RGB color model; therefore, the analysis shows that there is a statistically significant difference (p < 0.05) between groups. It can be concluded that Tequila 100% agave Cristalino is a Tequila 100% agave aged or extra-aged without color in which its chromatographic and isotopic profile is not affected. Full article

Achieving global sustainable agriculture requires farmers worldwide to adopt smart agricultural technologies, such as autonomous ground robots. However, most ground robots are either task- or crop-specific and expensive for small-scale farmers and smallholders. Therefore, there is a need for cost-effective robotic platforms that are modular by design and can be easily adapted to varying tasks and crops. This paper describes the hardware design of a unique, low-cost multiaxial modular agricultural robot (ModagRobot), and its field evaluation for soybean phenotyping. The ModagRobot’s chassis was designed without any welded components, making it easy to adjust trackwidth, height, ground clearance, and length. For this experiment, the ModagRobot was equipped with an RGB-Depth (RGB-D) sensor and adapted to safely navigate over soybean rows to collect RGB-D images for estimating soybean phenotypic traits. RGB images were processed using the Excess Green Index to estimate the percent canopy ground coverage area. 3D point clouds generated from RGB-D images were used to estimate canopy height (CH) and the 3D Profile Index of sample plots using linear regression. Aboveground biomass (AGB) was estimated using extracted phenotypic traits. Results showed an R², RMSE, and RRMSE of 0.786, 0.0181 m, and 2.47%, respectively, between estimated CH and measured CH. AGB estimated using all extracted traits showed an R², RMSE, and RRMSE of 0.59, 0.0742 kg/m², and 8.05%, respectively, compared to the measured AGB. The results demonstrate the effectiveness of the ModagRobot for in-row crop phenotyping. Full article

In recent years, biosensors have emerged as a promising solution for therapeutic drug monitoring (TDM), offering automated systems for rapid chemical analyses with minimal pre-treatment requirements. The use of saliva as a biological sample matrix offers distinct advantages, including non-invasiveness, cost-effectiveness, and reduced susceptibility to fluid intake fluctuations compared to alternative methods. The aim of this study was to explore and compare two types of low-cost biosensors, namely, the colourimetric and electrochemical methodologies, for quantifying paracetamol (acetaminophen) concentrations within artificial saliva using the MediMeter app, which has been specifically developed for this application. The research encompassed extensive optimisations and methodological refinements to ensure the results were robust and reliable. Material selection and parameter adjustments minimised external interferences, enhancing measurement accuracy. Both the colourimetric and electrochemical methods successfully determined paracetamol concentrations within the therapeutic range of 0.01–0.05 mg/mL (R² = 0.939 for colourimetric and R² = 0.988 for electrochemical). While both techniques offered different advantages, the electrochemical approach showed better precision (i.e., standard deviation of response = 0.1041 mg/mL) and speed (i.e., ~1 min). These findings highlight the potential use of biosensors in drug concentration determination, with the choice of technology dependent on specific application requirements. The development of an affordable, non-invasive and rapid biosensing system holds promise for remote drug concentration monitoring, reducing the need for invasive approaches and hospital visits. Future research could extend these methodologies to practical clinical applications, encouraging the use of TDM for enhanced precision, accessibility, and real-time patient-centric care. Full article